Two-week observation of Chinese AI: hardware shocks, software pessimism, founders surprised

Author: José Maria Macedo, Co-founder of Delphi Labs; Translated by: Golden Finance Claw

I spent two weeks in China meeting founders, venture capitalists (VCs), and CEOs of publicly listed companies across the entire AI ecosystem. Before going, I was optimistic about this ecosystem, expecting to find world-class AI geniuses starting companies at a fraction of Western valuations.

By the time I left, my perspective had become more nuanced: I was more confident in hardware than I had anticipated, more pessimistic about software, and some views on Chinese founders surprised me.

Founder Issues

The great founders I’ve invested in all share highly recognizable traits: independent thinking, rebelliousness, extreme focus, and near obsessiveness. They don’t follow orders, constantly ask “why,” and refuse to accept conventional norms. Their decisions may seem incomprehensible to outsiders, but to themselves, they are obvious. They possess a deep-seated, relentless strength, often reflected in a long-term obsession with excellence. Their lives have a certain “edginess” that makes them stand out among the many high-IQ talents I’ve met in VC.

However, many Chinese founders I encountered fit a different archetype — which surprised me.

They are talented — graduates from top universities, former employees at ByteDance or DJI, published papers in Nature, hold multiple patents. Achievements that only top-tier technical geniuses in the West would typically have are here just “entry tickets.” They are also some of the hardest-working people I’ve seen. We met at any time, on weekends, in different cities. One founder even came to see us on the day his wife was giving birth!

Yet, the qualities of independent thinking, rebellious spirit, and long-term vision from 0 to 1 are hard to find. These founders tend to have similar backgrounds, their business plans lean toward risk mitigation, and their ideas are often “impressive V2 versions” of existing things rather than truly original bets. Given the scale of technical talent in China, I expected to see more novel, unheard-of ideas.

My view is that China’s education system produces excellence but leaves little room for “deviation from the norm.” As a result, these founders are brilliant at solving known problems, but not the kind who can identify “problems no one knows about.”

VCs Reinforcing This Pattern

Even more interestingly, local investors are actively reinforcing this pattern.

Most Chinese funds’ investment logic revolves entirely around backing elite alumni from ByteDance or DJI — valuing backgrounds over edge, credentials over conviction. The VC community itself reflects this: most come from big corporations, consulting, or investment banking, similar to European VCs a decade ago.

Ironically, historically, China’s most outstanding founders — those who built epoch-defining companies — have never worked at large corporations. Jack Ma was an English teacher who failed the college entrance exam twice; Ren Zhengfei founded Huawei after leaving the military at 43; Liu Qiangdong started with a stall in Zhongguancun; Wang Xing dropped out of university and started his business from day one. A recent example is Liang Wengfeng, who, before founding DeepSeek, had never worked anywhere else besides his own company. These are “outliers,” lacking the so-called credentials — precisely the profile the current system tends to overlook.

There is genuine alpha (excess return) in finding talent with this profile, but few seem to be paying attention.

Shenzhen and the Hardware Ecosystem

What shocked me most in China wasn’t the startup pitch decks.

It was Shenzhen’s underground hardware scene — workshops where engineers systematically acquire Western high-end products, disassemble components, and reverse-engineer with extreme precision. By the end, I was genuinely unsure whether most Western hardware founders understand what kind of opponent they are competing against. The network effects here are not theoretical but physical, dense, and built up over decades.

Data from the companies we met confirmed this: over 70% of hardware investments come from the Greater Bay Area, nearly 100% from China itself — enabling iteration cycles that Western hardware companies simply cannot match.

Most founders I met follow the “DJI model”: building consumer-grade hardware in niche markets (like electric wheelchairs, robotic lawnmowers, next-gen fitness devices), reaching 8-9 figures in revenue, then expanding into adjacent categories using their customer base or underlying technology. Some of these businesses are already far beyond what I imagined. The most impressive company I encountered was Bambu, a 3D printing firm most Westerners haven’t heard of, reportedly earning $500 million annually, doubling every year.

Pessimism Toward Chinese Software

My doubts about Chinese software opportunities grew even more than when I arrived.

At the model layer, China’s open-source models are impressive — but closed-source models still lag significantly behind the top Western ones, and the gap may widen. CapEx differences are huge, GPU access remains limited. Western labs are increasingly restricting model distillation. Revenue data clearly illustrates the issue: reports say Anthropic alone did $6 billion in February, while China’s best models’ annual recurring revenue (ARR) is only in the tens of millions.

In software startups, the mainstream profile is former ByteDance product managers and researchers building agent or environment perception consumer apps for Western markets. Talent exists, but many products are within the scope of big tech’s native features — once a big company releases a version, these products become redundant. I was also struck by the lack of large, fast-growing private software companies. In the West, aside from model companies, several startups (like Cursor, Loveable, ElevenLabs, Harvey, Glean) are generating 9- or 10-figure ARR with astonishing growth. Such breakout private software companies are almost nonexistent in China — with few exceptions like HeyGen, Manus, and GenSpark, which all eventually chose to leave once opportunities arose.

Valuation Bubble

Despite the promising software outlook, bubbles in early and late stages are very real.

In early stages, although top talent from ByteDance, DeepSeek, and Moonshot remains cheaper than comparable US talent, median valuations are converging. It’s common for consumer startups to be valued at $100–200 million before product launch. Seed pre-funding rounds exceeding $30 million are not rare.

In later stages, the numbers are even harder to justify. Minimax’s public market valuation is about $40 billion, with ARR under $100 million — roughly 400x sales. Zhizhi (智谱) is valued at around $25 billion with $50 million in revenue. By comparison, OpenAI’s peak funding round had a price-to-sales ratio of about 66, Anthropic around 61.

Non-listed companies like Moonshot leverage these public comparables for funding, raising $6 billion, $10 billion, and $18 billion in just a few months. Crypto investors recognize this pattern: they compare private valuations to ununlocked fair market prices. Part of the support for Zhizhi and Minimax at these levels is that they are currently the only ways to gain exposure to the “Chinese AI narrative,” thus carrying a premium. As more companies go public and this trait dilutes, the situation will change. IPO windows tend to close quickly and without warning — you can’t be sure you’ll complete arbitrage before the benchmark prices shift.

The humanoid robot sector is similar. China has about 200 humanoid robot companies, with around 20 raising over $100 million, several valued at billions — almost all with no revenue, most planning Hong Kong IPOs in 2026 or 2027. If this market is real, China’s hardware dominance makes long-term outcomes quite clear. But commercialization may be slower than current funding rhythms suggest, and I doubt Hong Kong can support so many billion-dollar humanoid robot companies. I am currently watching from the sidelines.

The Asymmetry Worth Watching

One thing I didn’t expect: nearly every founder I met is building products for the global market, not just for China. They use Claude Code, watch Dwarkesh’s videos. They are well-versed in the San Francisco startup scene, often more than Western investors who haven’t paid close attention.

Western hostility toward China is far deeper than China’s toward the West. Chinese founders see no contradiction in combining China’s engineering execution, hardware depth, with Western go-to-market strategies and product visions. When this combination appears in the right founding team, it can produce truly outstanding companies.

Finding these founders — outliers who don’t fit the local VC ecosystem’s “elite credentials” mold — is exactly what we are focusing on now.